RFI's and so forth

This being my 4th or 5th post in just over a week, I will make an 
attempt at brevity.

What you want to do is possible, but difficult.  First, of all, I'd 
like to refer you to what I consider the seminal paper on this topic: 
Roederer, M. and Herzenberg, L. A. (1996). Changes in antigen 
densities on leukocyte subsets correlate with progression of HIV 
disease. Int Immunol 8, 1-11.

(Please, the fact that the first author on this paper shares my last 
name and first initial should in no way affect your view of my 
assessment of the seminality of this paper).

In this paper, the RFI of over a dozen different antigens were 
compared across collections over a period of many months.  There was 
a high degree of success, but this was driven by several important 
factors:  (1) the same lot of reagents were used across all 
experiments; (2) a very consistent calibration scheme was employed. 
These are not necessary, but significantly improved the reduction in 
the CV across collections.

In your case, where no consistent calibration standard was chosen, 
your only option is the "delta-delta-RFI" that was suggested by 
someone whose email I have unfortunately deleted already.  However, 
this approach will only work if your basis measurement is something 
that is highly conserved (in other words, if you base your 
delta-delta-RFI on an antigen which is itself varying, then you are 
not going to have success).  It has been suggested that the CD4 
fluorescence falls into this category, as the number of CD4 molecules 
per CD4 T cell is highly regulated.

But if you switch lots of reagents, your only hope is to perform a 
validation across the lots--you will have to have measured the same 
sample with the different lots of reagent to be able to come up with 
a correction factor.

What you begin to see is that you will have several correction 
factors for each collection:  (1) normalizing the fluorescence to a 
given "reference"; (2) normalizing this value by the lot-to-lot 
correction.  At some point, the errors in the corrections will far 
outweigh the actual deviations that are meaningful.

(Let's also not forget the all-important question:  did you use the 
reagents at saturating titres?	If you did not titrate the reagents, 
then you don't know the answer to this question.  And if you don't 
know the answer, then don't even bother trying to compare RFI's. 
They could depend on staining temperature, staining time, and (as 
recently discussed), in some cases, on the number of cells in the 
sample.  (There, Joanne, are you happy?).  Bottom line: if your PI 
did not titrate the reagents, then tell the PI that they are out of 
luck; no chance).

Finally, I'd like to also say that the fact that the "background 
fluorescence is pretty darn consistent" should NOT be taken to mean 
that your measurements are consistent.	Background fluorescence is 
often a measurement made on a few photoelectrons, and can largely 
consist of electronic noise--so the fact that it is consistent should 
in now way make you feel comfortable that your real fluorescence 
measurements are in any way meaningfully comparable.  The only way to 
judge this is to compare some potentially conserved measurement 
(e.g., CD4) -- but how will you know in advance that it is consistent 
to begin with?

I suggest that there are too many potential problems in analyzing 
your data to get comparable RFI values.  Certainly, a reviewer would 
have a field day with any such presentation.  However, if you want to 
look at it to gather some ideas for future analysis, then by all 
means go for it.  For future analysis, make sure that (1) you use 
reagents in saturation; (2) you have a fluorescence standard; (3) you 
calibrate the machine to consistent sensitivities; (4) you can either 
use a single lot of reagent or perform cross-lot validation; and (5) 
you validate the whole concept before you actually start.

OK, time to step down from my soap box.  (Sorry, it's a long way 
down).	I realize I have failed at my misearable attempt at brevity, 
but perhaps not at levity...  in the future, I will try not to write 
these lectures in a post-prandial state.

mr

>I rec'd several replies to my last post re:comparing RFI's, but none 
>of them addressed my actual question.
>I will try to clarify my question:
>
>We have data for a particular study that goes back over a year. In 
>the beginning, the PI was only interested in % expression. Each 
>subject had multiple experimental controls (stimulated, unstim , 
>stained & unstained, etc) so the measurement of % increase/decrease 
>expression is not the issue.
>
>Several months ago, the question came up to compare the RFI's of the 
>expression. Well, not having included any calibration-type bead (or 
>other control) I didn't feel you could take separate experiments and 
>compare the RFI's across the board. (Since that time, calib beads 
>are run with every experiment).
>Like most labs, the instrument has been adjusted by service 
>engineers several times. Each run is also slightly different in 
>settings.  Since an unstained control (N-1 sample) was included in 
>each run, I was wondering if it was possible to somehow adjust the " 
>scale" (or whatever) for each subject ,so that ALL the samples of 
>ALL subjects  could be compared for relative fluorescent intensity.
>
>We aren't trying to quantify number of molecules or anything  like 
>that. I'm curious if it is valid to compare , for example: all 
>treatment group1 of all 125 subjects, even though they were run 
>slightly different.
>The curious thing is that when we compared the RFI's, they actually 
>fell into some interesting patterns (with respect to the "groups"). 
>The background fluorescence for the N-1 in the channel of interest 
>is pretty darn consistent, so the argument I was presented with is 
>"Why not?"
>
>On a slightly different twist: for those subjects that we now have 
>"calibration" standard data for the instrument (during data 
>collection), is it even valid to compare the RFI's if different LOTS 
>of antibodies are used?  I understand the use of MESF - qualified 
>Ab's for quantification of receptors & such.  If you don't use 
>qualified Ab's , are RFI's only valid within a particular experiment?
>
>Thanks!
>
>I will post a summary- I had many folks interested in the answer!
>
>bunny

-- 
_____________________________________________
Mario Roederer, Ph.D.
Chief, ImmunoTechnology Section and Flow Cytometry Core
Vaccine Research Center, NIAID, NIH
40 Convent Dr., Room 5509
Bethesda, MD 20892-3015
Phone: 301 594-8491
FAX: 301 480-2651